Purpose: High‐Z material in CT yields metal artifacts that degrade image quality and may cause substantial errors in dose calculation. This study couples a metal artifact reduction (MAR) algorithm with enhanced 16‐bit depth (vs. standard 12‐bit) to quantify potential dosimetric differences. Methods: Extended CT to electron density (CT‐ED) curves were derived from tissue characterization phantom with Titanium and Stainless Steel inserts scanned at 90–140 kVp for 12 and 16‐bit reconstructions. MAR was applied to sinogram data (Brilliance BigBore CT scanner, Philips Healthcare, v.3.5). Single/double hip prostheses were simulated using Cerrobend rod(s) embedded in a pelvic phantom and two dosimetric verifications were performed to elucidate changes between 12 and 16‐bit data: (1) BEAMnrc/Dosxyz Monte Carlo simulation (MC‐SIM) (400,000,0000 histories, 6×, 10×10 beam traversing Cerrobend rod) and (2) Gafchromic film analysis for plans calculated with Anisotropic Analytic Algorithm (AAA, Eclipse) and Monte Carlo (MC‐BL, Brainlab). Five patients with metal implants were reconstructed using both bit depths, with plans calculated using AAA and derived CT‐ED curves. Dose profiles and matrices were evaluated. Results: For 12‐bit images, Titanium and Stainless saturated at 3071 HU, while for 16‐bits (120kVp), mean CT numbers were 8088+/−336 and 13,971+/−1460 HU for Titanium and Stainless, respectively. MC‐SIM was well‐matched between 12 and 16‐bit images except downstream of metal, where 16‐bit dose was ∼6.4% greater than 12‐bit. In treatment planning phantom cases, dose profiles were well‐matched in the buildup region, but deviated in the shadow of metal where ∼12% differences between AAA 16‐bit and 12‐bit doses were observed. Similar results were obtained in patient cases. The largest discrepancy was at a tissue‐metal prosthetic interface, with 12‐bit dose yielding 23.8% less dose than 16‐bit. Conclusions: Dosimetric differences revealed between 12‐ and 16‐bit images were substantial. Further dosimetric verification is warranted. These results support implementing MAR‐corrected, 16‐bit images in treatment planning. This work supported in part by a grant from Philips Medical Systems, Cleveland. Titanium and Stainless Steel inserts were provided by Gammex Inc. in exchange for 16‐bit depth information.